TW201003651A - Method of making record on magnetic memory device - Google Patents

Abstract

A method of making a record in a magnetic memory device which has a memory layer which holds information as magnetization directions of a magnetic material and a magnetization reference layer which is provided over the memory layer with an insulating layer interposed therebetween. Recording is made by using current flowing between the memory layer and the magnetization reference layer through the insulating layer. An error rate as high as that of when a write pulse a little higher than the inversion threshold value is applied can be maintained even when a write pulse considerably higher than the inversion threshold value is applied. To record one piece of information, at least one main pulse and at least one subpulse are applied in the same direction. The main pulse is a pulse having a pulse height and a pulse width both of which are great enough to record the information. The subpulse is a pulse satisfying at least one of two conditions: one that the pulse width thereof is shorter than that of the main pulse and the other that the pulse height thereof is lower than that of the main pulse. At least one subpulse is applied after the main pulse is applied.

Description

201003651 VI. Description of the Invention: [Technical Field] The present invention relates to a magnetic memory element ^ t β recording method, the magnetic property. The body element contains a memory layer, and the magnetization direction of the 彳μ 士八系 can be changed, and the information is held as the magnetization direction of the magnet; and the magnetic IU丞+ layer is provided with the isolation edge layer, which is the magnetization direction. The basis ' 曰 „ ' and the information is recorded by flowing through the insulating layer between the memory layer and the magnetization reference layer.

[Prior Art] In a computer such as a random access memory (10) nd (10)

Access Memory (Random Access Memory) is a DRAM (Dynamie RAM: Dynamic Continuation) that uses high-speed operation and high-density recording. However, since the D pair is a volatile memory that disappears after the power is turned off, it is strongly expected that the information can be kept and the non-volatile memory that is indispensable for the low power consumption of the device can be speeded up even if the power is turned off. And high density and large capacity. ^ As a non-volatile memory, flash memory, etc. have been put into practical use. In recent years, as a non-volatile memory with high speed, large capacity, and low power consumption, magnetic memory using magnetic resistance effects has been developed. progress. For example, a magnetic memory element using a tunnel magnetoresistance (TMR) effect, that is, an MTJ element, is used to invert the magnetization direction of the memory layer by using a magnetic field excited by a current. The magnetic random access memory (Magnetic ram: MRAM) is used only (for example, MR2A1 6 (137416.doc 201003651 trade name) manufactured by Freescale Semiconductoi Co., Ltd.). Fig. 9(a) is an explanatory view showing the basic structure of the MTJ element and the read operation of the memory information. As shown in FIG. 9(a), the MTJ element 1 has a so-called magnetic tunnel junction (MTJ), which is applied to two kinds of ferromagnetic layers of the sensible layer 1〇5 and the magnetization reference layer 1〇3. Between, the structure of the non-magnetic thin insulating layer, that is, the channel insulating layer 1〇4 is sandwiched. The memory layer 1〇5 is composed of a strong magnetic conductor having uniaxial magnetic anisotropy, which can change the magnetization direction by external action and maintain the magnetization direction as information. For example, the magnetization direction is "parallel" or "anti-parallel" to the magnetization direction of the magnetization reference layer 1 〇 3, and is used as the information memory of "0" and r 丨 respectively. The information reading system of the MTJtl device uses the T]y[R effect, which is due to the difference in the relative magnetization directions of the above two magnetic layers, and flows to the memory layer 105 and the magnetization through the channel insulating layer 1〇4. The resistance value of the channel current between the reference layers 1〇3 changes. This resistance value is obtained by taking the minimum value when the magnetization direction of the memory layer 1〇5 is parallel to the magnetization direction of the magnetization reference layer 103, and the maximum value when the anti-parallel is performed. Figure 9(b) shows the MTJ component! 00 is a partial perspective view of the structure of the memory cell of the iMRAM. In the Mram, the word line as the column wiring and the bit line as the row wiring are arranged in a matrix, and the MTJ element 1 is placed at the position of each of the intersections to form a memory equivalent to 1 bit. Body cell. In the upper part of the cell, the write bit line i22 and the read bit line 123 are provided with the interlayer insulating film interposed therebetween, and the MTJ element 1 is connected to the master bit line 123. And placed underneath, the input word line 12ι is placed next to the electrode layer 106 under the electrode layer 106 of the MTJ element 1 '. On the other hand, in the lower portion of the memory cell, a MOS (Metal Oxide Semiconductor) field effect transistor is provided on the semiconductor substrate 111 such as a germanium substrate, for reading. Select during the action. The memory cell is selected by the transistor 110. Gate electrode of transistor 110

The n5 system is connected to each other to form a strip shape, which also serves as a reading word line. Further, the source region 114 is connected to the MTJ^; 70-piece lead-out electrode layer via the read-out connection plug 1〇7; the drain region 116 is connected to the sense of the wiring for reading. Line 124. In the MR AM thus constructed, the write (record) of the MTj element 100 of the desired memory cell is written by the write word line 12 and the row respectively included in the memory cell. The write bit line 122 flows into the position where the write current 'at the intersection of the two write wirings, and the combined magnetic boundary of the magnetic boundary caused by the currents is generated. By means of the synthetic magnetic field, the memory layer 105 of the component (10) is desirably stored in the direction of the specific magnetization U 7 ', that is, in the direction of "parallel" or anti-flat" to the magnetization direction of the magnetization reference layer (8). Magnetization, writing (recording) information. - The information from the MTJ element 100 is read from the information element containing the desired one, and the 5th output sub-line of the I, that is, the gate electrode π 5 is applied to select L唬. The selected transistors 1 1 〇 are all turned on (on). The read voltage is applied between the read bit line 123 and the sense line 124 in the row including the desired memory cell. As a result, only the desired memory cell is selected, and the difference in the magnetization direction of the memory layer 1〇5 of the MTJ element is used as the channel current 137416.doc 201003651 to the MTJ element 1 using the 丽 effect. It is detected by omitting the difference in size. The channel band is measured by the peripheral circuit in the case where the circuit L is taken out from the sensing line 124. The non-volatile memory of the TMR type, which uses the magnetic resistance effect according to the spin-dependent conduction phenomenon unique to the Nai (four) body, is rewritten by the reversal of the magnetization direction, so Unlimited rewriting can be performed 'About access time is also reported as high speed (see for example R. Scheuerlein et al τςςί-r η 5 iSSCC DlSest Technical Papers, pp. 128-129, Feb_2000). However, in the case of writing a person's temple with a current magnetic field, a large amount of current (for example, a few claws) has to be flown for rewriting, and power consumption is increased. Further, when the MTJ element is miniaturized, the current necessary for rewriting exhibits an increased tendency. In addition, since the write wiring is thinned, it is difficult to flow a current sufficient for rewriting. Moreover, the high integration progresses, and the accuracy of mis-writing of other memory cells adjacent to each other becomes high. Further, since the writing wiring and the reading cloth and the line are separately required, the structure is complicated. For these reasons, (4) the high density and large capacity of the MRAM to be written by the current magnetic boundary. Therefore, as an element for writing (recording) information to a memory layer of a magnetic memory element according to a different principle, a magnetic memory element which utilizes magnetization reversal by spin injection at the time of writing is attracting attention. The spin injection generates a current composed of an electron group that is biased toward one side by a direction of spin (spin_p〇larized current) by a current flowing in a strong magnetic conductive layer (magnetization reference layer) in which the magnetization direction is fixed. And the current, the operation of the magnetic conductive layer (memory layer) which is mainly changed in the magnetization direction. In this way, when the spin-polar current flows in the memory layer, the magnetization of the memory layer is caused by the action of the spin-polarized electric 137416.doc 201003651 and the electrons of the magnets constituting the memory layer. The direction and the magnetization direction of the magnetization reference layer - the force (torque) of the 会 will work. Therefore, the magnetization direction of the memory layer can be made by the spin-polar motor that flows into the current density above +^ of a certain threshold. The anti-Korean-Chinese transfer (see, for example, Patent Document 1 and Non-Patent Document 1 to be described later).

Fig. 1 shows the magnetization of the MTJ element (hereinafter referred to as the self-injection and the element) which is formed by the inversion of the magnetization square shown in the second reading, and the magnetization caused by the spin injection. A partial perspective view of the construction of the difficulty of reversing (hereinafter referred to as spin torque MRAM). In the spin torque MRAM, a word line 215 as a column wiring and a bit line 218 as a row wiring are arranged in a matrix, and one spin injection MTJ element 220 is disposed at a position of each of the intersections. A memory cell corresponding to the 丨 bit is formed. Figure 10 shows 4 parts of memory cells. On the lower semiconductor substrate 211, a selection transistor 21, which will be described later, is formed in each memory cell, and the word line 215 also serves as a gate electrode of the selection transistor 21A. Further, the drain region 216 is formed in common in the left and right selection transistors in the figure, and the column wiring 219 is connected to the drain region 216. Figure 11 is a partial cross-sectional view showing the structure of a memory cell of a spin torque MRAM. In the central portion of the memory cell, a base layer 20 1 , an antiferromagnetic layer 202 , a magnetization fixed layer 203 a , an intermediate layer 203 b , a magnetization reference layer 203 c , a channel insulating layer 204 , a memory layer 205 , and protection are sequentially laminated from the lower layer. Each layer of layer 206 is formed with a spin implant MTJ element 220. The layer composition of the spin-injected MTJ element 220 is substantially the same as that of the conventional MTJ element. The magnetization fixed layer 203a, the intermediate layer 203b, and the magnetization reference layer 203c are laminated 137416.doc 201003651 on the antiferromagnetic layer 202 to constitute a fixed magnetization layer as a whole. The magnetization direction of the magnetization fixed layer 203a composed of a strong magnetic conductor is fixed by the antiferromagnetic layer 202. Similarly, the magnetization reference layer 2〇3() composed of a strong magnetic conductor forms an antiferromagnetic bond with the magnetization fixed layer 203a via the intermediate layer 203b of the nonmagnetic layer. As a result, the magnetization direction of the magnetization reference layer 203c is fixed in the opposite direction to the magnetization direction of the magnetization solid layer 203a. In the example shown in Fig. 11, the magnetization direction of the magnetization fixed layer 203a is fixed to the left, and the magnetization direction of the magnetization reference layer 2?3c is fixed to the right. If the solid-state magnetization layer is formed into the laminated iron structure described above, the sensitivity of the fixed magnetization layer to the external magnetic boundary can be reduced. Therefore, the magnetization variation of the fixed magnetization layer due to the external magnetic boundary can be suppressed, and the stability of the MTJ element can be improved. . Further, since the magnetic fluxes leaked from the magnetization fixed layer 2〇3a and the magnetization reference layer 2〇3c cancel each other, the magnetic flux leaking from the fixed magnetization layer can be suppressed to the minimum by adjusting the film thicknesses. Z fe and layer 5 are composed of a strong magnetic conductor having uniaxial magnetic anisotropy, and the magnetization direction can be changed by external action, and the magnetization direction can be maintained as a poor signal. For example, the magnetization direction is "parallel" or "anti-parallel" to the magnetization direction of the magnetization reference layer 2〇3c, and is used as the information memory of "〇" and "1", respectively. Between the magnetization reference layer 20 氕 and the memory layer 2 〇 5, a channel insulating layer 2 〇 4 having a non-magnetic thin insulating layer is further disposed by the magnetization reference layer 203 c, the channel insulating layer 2 〇 4 and the memory layer 2 〇5 forms a magnetic channel junction (MTJ). On the other hand, in the lower part of the memory cell, the well region 2 i 1 & in the plate 2 1 1 is separated from the semiconductor substrate of the ruthenium substrate or the like as the memory for selecting the memory 137416.doc 201003651 The selection transistor 210 is provided with a gate insulating film 2 12 , a source private electrode 2 13 , a source region 2 丨 4 , a gate electrode 2 丨 $ , a drain region 2 丨 6 , and a drain electrode 21 7 . The m〇s type field effect transistor is composed. The gate electrode 21 of the transistor 210 is connected to each other to form a strip shape, and also serves as a word line for the second column wiring. Further, the electrodeless electrode 217 is connected to the column wiring 219 which is the second column wiring, and the source electrode 213 is connected to the base layer 2〇1 of the spin injection MTJ element 220 via the connection plug 2〇7. On the other hand, the protective layer 2〇6 of the spin-implanted MTJ element 220 is connected to the bit line 2 18 which is provided as a row wiring on the upper portion of the memory cell. When the spin of the desired memory cell is injected into the MTJ element 22, the selection signal is applied to the word line 2丨5 including the column of the desired memory cell, so that the selection transistor 21 of the column is all To be on (on) state. The write voltage is applied between the bit line 2 18 and the column wiring 219 including the row of the desired memory cell. As a result, the desired memory cell is selected, and the spin-polar current flows through the memory layer 205 of the spin-implanted MTJ element 22, and the memory layer 2〇5 is magnetized in a specific magnetization direction to perform information recording. At this time, first, the magnetization direction of the magnetization reference layer 2〇3c of the spin injection MTJ element 22 is "anti-parallel" to the magnetization direction of the brother layer 2 〇5, and is reversed by writing to When the magnetization direction of the memory layer 2〇5 is in a state of “parallel” to the magnetization direction of the magnetization reference layer 2〇3c, as shown in FIG. 11, the write current of the current density above the threshold value flows from the memory layer 205. To the magnetization reference layer 2〇3c. Thereby, as an entity, the spin-polar electron flow of the electron density above the threshold 137416.doc 201003651 flows from the magnetization reference layer 203c to the memory layer 205, causing magnetization reversal. On the other hand, when the magnetization direction of the magnetization reference layer 203c in which the magnetization direction of the memory layer 205 is in the "parallel" state is reversed to the "anti-parallel" state, the current density of the current density above the threshold value is applied to the above. In the opposite direction, that is, from the magnetization reference layer 203c to the memory layer 205, as an entity, an electron flow of electron density above the threshold flows from the memory layer 205 to the magnetization reference layer 203c. Further, the information reading from the spin injection MTJ element 220 is performed in the same manner as the MTJ element 100 by the TMR effect. Both the writing and the reading of the spin-injected MTJ element 220 utilize the interaction of the electrons in the memory layer 205 with the spin-polar current flowing through the layer, and the current density of the spin-bias current is small. The area is written in a region where the current density of the spin-bias current is large and exceeds the threshold. Since the magnetization reversal by spin injection depends on the current density of the spin bias current, the spin of the MTJ element 220, the smaller the volume of the memory layer, proportional to the volume, and less The current can be reversed by magnetization (refer to Non-Patent Document 1). Further, since the memory cell selected by the transistor 210 is selected to write information, unlike the writing by the current magnetic field, there is no possibility that another adjacent cell is mistakenly written. Moreover, most of the wiring can be shared by writing and reading, so that the structure is simplified. Further, since the influence of the shape of the magnet is smaller than that of the magnetic field writing, it is easy to improve the yield at the time of manufacture. From these viewpoints, the spin torque MRAM is more suitable for the fine-grained MRAM than the MRAM that is written by the current magnetic field. However, since writing (recording) is performed using the selection transistor 210, other problems are caused. That is, the current which can flow through the spin injection MTJ element 220 at the time of writing is limited by the current (saturation current of the electric crystal) which can flow through the selection transistor 210. In general, as the gate width of the transistor or the gate length becomes smaller, the saturation current of the transistor also becomes smaller. Therefore, in order to ensure the write current for the spin injection of the MTJ element 220, the small size of the transistor 210 is selected. Limited access. Therefore, in order to miniaturize the selection transistor 210 r and ' miniaturization, the spin torque MRAM is maximized and increased in capacity, and it is indispensable to reduce the threshold of the write current as much as possible. Moreover, in order to prevent insulation damage of the channel insulating layer 204, it is necessary to reduce the threshold of the write current. Moreover, in order to reduce the power consumption of the MRAM, it is also necessary to reduce the write current threshold as much as possible. The threshold value of the current required for the magnetization reversal by the spin injection shows a quadratic square of the spin-braking constant α and the saturation magnetization Ms of the memory layer 205. And volume V proportional, and spin injection efficiency

K.J η is inversely proportional. Therefore, by appropriately selecting these, the threshold of the current required for magnetization reversal can be reduced. On the other hand, however, in order to become a reliable memory element, the spin injection MTJ element 220 must ensure the memory retention characteristics (thermal stability of magnetization) of the memory layer 205, and the magnetization direction does not change due to thermal motion. The thermal stability is proportional to the saturation magnetization Ms and the volume V of the memory layer 205. The saturation magnetization Ms and volume V of the memory layer 205 are related to both the threshold and the thermal stability of the current required for magnetization reversal, and are in the process of reducing the 137416.doc -11 - 201003651 factor, so that the magnetization reversal + ^ The trade-off relationship is also reduced.疋Nursing: Face Barrier::: The threshold of the current required for magnetization reversal is reduced, which must be established at the same time as the thermal stability is ensured. η. In order to make the spin force MRAM become a more distracting memory than other 疋 立 , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , (4) Established MTW material (see Japanese Patent Laid-Open No. 2GG6-165265, JP-A-7) (4)... Tiger Gazette, Japanese Special Report No. 7_48. Patent Document 2 and Japanese Patent Application No. 2006-35G113, etc.). As a result, it is constantly approaching its implementation. [Patent Document 1] Japanese Patent Laid-Open Publication No. Hei 2〇〇3_m82 (Sections s and 7 and Fig. 2 [Patent Document 2] Japanese Laid-Open Patent Publication No. 7-287923 (page 7-15, Fig. 2) Patent Document 1] FJAlbert et al., Appl. Phys. Lett., v. 1 77, (2002), p. 3809 [Disclosure] [The problem to be solved by the invention] However, the inventor of the present invention uses the MTj material described above. Produce and investigate the spin injection MTJ component with a small threshold value of the current density, and identify the special phenomenon that has not been reported in the paper or the publication of the society. That is, the MTJ component is injected into the spin to confirm Even if the write error rate is taken into consideration, the applied write pulse wave setting is slightly larger than the inversion threshold value (as an estimated value obtained by extrapolation), and the write error rate of 10-25 or less can be secured, but 137416.doc •12- 201003651 If the added pulse is set to be greater than the inversion threshold, the write pulse will be larger, but the write error will increase (refer to the figure). Therefore, the error caused by the recording voltage greater than the inversion threshold is called "high." Recording voltage error. "

γ - the spin torque mram memory chip with hundreds of Mbits, which is based on the deviation of the inversion threshold of the spin injection MTJ component or the reversal of the transistor and wiring The deviation of the threshold value, etc., is set as the write pulse wave whose target is added to be larger than the average value of the inversion threshold. Therefore, if the above phenomenon occurs in the actual writing of the spin torque MRAM memory chip, the write error rate of 1 〇 25 or less cannot be ensured. Further, since the MRAM or the spin torque RAM maintains information as the magnetization direction of the magnet constituting the memory layer, if it is exposed to a strong external magnetic field, the magnetization direction of the memory layer changes, and the information disappears. Especially in the middle of the writing (recording) process, the tolerance to the external magnetic field is significantly reduced: therefore, a magnetic shield for reducing the external magnetic field acting on the magnetic memory element is required, which is also available in the aforementioned commercially available 2MRAM (MR2A16). Equipped. However, in order to obtain a magnetic field shielding effect by magnetic shielding, a certain degree of thickness and volume is required, and it is impossible to avoid the increase in the volume or weight f of the memory IC, and the increase in the = cell.

Different from the spin torque RAM, for example, in the literature (Kjt〇et a丨., DV〇l. 40, 2007, p. 1261) indicates that there is an external magnetic field affecting the recording current or the inversion time, further in the literature (GD Fu café et al., Aplphys Ut, P.1525G9) indicates that there is a possibility of spin-injection of the MTT component due to energization, and the tolerance to the external magnetic field is further reduced by 137416.doc 201003651. Improve the tolerance to external magnetic fields. The present invention has been made in view of such circumstances, and an object thereof is to provide a recording method of a magnetic memory element, the magnetic memory element comprising: a memory layer which is changeable in magnetization direction and holds information as a magnetization direction of the magnet; And a magnetization reference layer, which is provided with respect to the memory layer through the isolation edge layer, and the reference of the direction, and the current flowing between the cell layer and the magnetization reference layer through the insulating layer, and recording information; Even in the case where a write pulse having a value greater than the inversion threshold is applied, the write error obtained in the case where the write pulse of = slightly larger than the inversion threshold is applied can be maintained. Wu rate and improved tolerance to external magnetic fields. [Technical means for solving the problem] The inventors have repeatedly conducted intensive studies, and as a result, found that the above problem can be solved by designing the writing method of wave application, and the present invention is completed. That is, the present invention relates to a recording method of a magnetic component, a magnetic memory component recording method, and a magnetic memory component: a memory layer comprising a strong magnetic conductor and a variable magnetization direction The word δ hole acts as a magnetization shot of the magnet && . is π guaranteed, and the reference magnetization layer, its ', child; month IJ memory layer dielectric isolation edge> The layer is s, and contains a strong magnetic conductor, the sound of the sound, the ώ Μ, the base of the +, and the current flowing through the aforementioned insulation increases the current between the quasi-magnetized layers of the fish front information , carry the shell. The rites of the rites are characterized by: m ^ when 1 has recorded 1 poor news, applying more than one main pulse wave and one brake pulse on m ^ in the same direction; more than one main pulse wave in the above After that, applying 4 or more pulses of the sub-m of the king of the king, the pulse wave applied after the main pulse wave is set to match the pulse wave t I ratio, and the pulse wave of the main pulse wave is short, or the pulse 137416. Doc 14 201003651 A pulse wave of at least one condition of a pulse wave lower than the aforementioned main pulse wave. Further, the aforementioned pulse wave may be voltage control, current control, or power control. [Embodiment] [Effect of the Invention] According to the recording method of the magnetic memory device according to the present invention, when the information is recorded as described in the following embodiments and examples,

After the main pulse wave, the plurality of sub-pulse waves are applied, and the sub-pulse wave applied after the main pulse wave is set to be a pulse wave having a pulse width wider than the main pulse wave, or the pulse wave height is higher than the main pulse wave. At least the pulse of the low pulse - even if the applied pulse wave is greater than the inversion threshold: the lower 'can still remain with the write pulse slightly larger than the reverse threshold The wave gets the same write error rate. Another mechanism for exhibiting a high recording voltage error as described above or a mechanism for suppressing the error rate of the horse by the present invention is not completely _. Lacking, from the application of the imitation is greater than the reversal threshold ..., does not constitute _, in the case of applying more than the inverse ... = wave in the case of the β B 孓夂 ring limit of the write pulse wave And the larger the write pulse, the more the write error rate, plus the thoughts, it can be speculated that compared to the reverse: cause problems. The injection of the incoming power is caused by a write error or the like caused by the force injection by the single pulse wave in the past. The write error rate is high due to the excess write power. Moreover, the original sample becomes the result. Therefore, in contrast to this, the tolerance to the external magnetic field is low. After the above one or more main pulse waves, apply! 137416.doc -15 - 201003651 More than the secondary pulse wave, so the writing by the secondary pulse wave can correct the possibility of writing errors caused by the main pulse wave. Further, since the sub-pulse wave applied after the main pulse wave is set to be a pulse wave having a pulse width wider than the main pulse wave, or a pulse wave having a pulse wave height lower than a pulse wave having a lower pulse wave than the main pulse wave, In the writing by the aforementioned secondary pulse wave, it is difficult to accumulate excess energy, and the above-mentioned high recording voltage error is unlikely to occur. With the above effects, the recording method of the magnetic memory element of the present invention has a 'write error rate reduction' and the writing is improved for the external magnetic field. In the method of recording a magnetic memory device according to the present invention, at least one set of consecutive 3s is included in the pulse train including the one or more main pulse waves and the one or more of the sub-pulse waves applied thereafter. A combination of pulse waves and a combination of at least one of the pulse waves and the pulse wave surface is gradually reduced. Further, a time interval of 3 ns or more is provided between the end of the one or more main pulse waves and the front end of the one or more of the sub-pulses described above. (In addition, the end of the pulse wave and the front end are respectively a position where the height of the pulse wave is one-half of the maximum value of the pulse wave during the fall and the rise of the pulse wave. The same applies to the above-mentioned "the main pulse wave of the above-mentioned one or more." In the pulse train of the above-mentioned one or more of the sub-pulse waves applied, the combination of the arbitrarily selected two consecutive pulse waves is preferably such that the pulse wave width is equal to or greater than 2 ns and less than 1 ns. Or a pulse wave having a pulse wave height of at least one of 7 times or more and 95 times or less of the front pulse wave, and a time interval of 5 ns or more between the end of the front pulse wave and the front pulse wave. Further, in the pulse wave train including the one or more main pulse waves described above and the subsequent application of 137416.doc • 16· 201003651, the pulse wave train of the plurality of pulse waves is arbitrarily selected. 'The posterior pulse wave should be at least one of the pulse width less than 3 ns, or the pulse wave height is less than 0.95 times the front pulse wave, and the time interval between the end of the front pulse wave and the front end of the back pulse wave is set to be less than 5 Ns. Next, a preferred embodiment of the present invention will be described in more detail with reference to the drawings. Implementation type 1

p is an implementation example 1, and is mainly directed to an example of a recording method of the spin injection MTJ element of the claims 1 to 3. Figs. 15 and 16 show the structure of the memory cell of the spin torque mram used in the present embodiment and the configuration of the spin injection MTJ element. Fig. 15 is a view showing an MRAM in which a magnetization reversal is performed by spin injection of a 2Mtj element (hereinafter referred to as a spin injection MTJ element), and magnetization inversion by spin injection is used (hereinafter referred to as self). Partial perspective view of one example of the rotational moment mram)2 configuration. In the spin torque mram, the word line 15 as the column Q wiring and the bit line 18 as the row wiring are arranged in a matrix, and one spin injection element is disposed at the position of each of the % of the parent points. , • Forms a memory cell equivalent to 1 bit. Fig. 15 shows four parts of memory cells. In the lower semiconductor substrate 丨1, a selection transistor 1 to be described later is formed in each of the memory cells, and the word line 15 also serves as a gate electrode for the selection transistor 1A. Further, the non-polar region 16 is selected from the left and right in the figure, and is commonly formed by the f crystal. The column wiring 19 is connected to the sigma-polar region 16. Fig. 16 is a partial cross-sectional view showing the configuration of the memory cell of the spin torque MRAM 137416.doc 201003651. In the central portion of the memory cell, a base layer 1 'antiferromagnetic layer 2, a magnetization fixed layer 3a, an intermediate layer 3b, a magnetization reference layer 3c, a channel insulating layer 4, a memory layer 5, and a protective layer are sequentially laminated from the lower layer. Each of the layers 6 is formed with a spin injection MTJ element 20. The magnetization fixed layer 3a, the intermediate layer 3b, and the magnetization reference layer 3c are laminated on the antiferromagnetic layer 2 to constitute a fixed magnetization layer as a whole. The magnetization direction of the magnetization fixed layer 3a composed of a strong magnetic conductor is fixed by the antiferromagnetic layer 2. Similarly, the magnetization reference layer 3c composed of a strong magnetic conductor forms an antiferromagnetic bond with the magnetization fixed layer 33 via the intermediate layer 3b of the nonmagnetic layer. As a result, the magnetization direction of the magnetization reference layer 3c is fixed in the opposite direction to the magnetization direction of the magnetization fixed layer 3a. In the example shown in Fig. 16, the magnetization direction of the magnetization fixed layer 3a is fixed to the magnetization direction of the leftward magnetization reference layer 3c to the right. When the fixed magnetization layer is formed into the laminated iron structure, the sensitivity of the fixed magnetization layer to the external magnetic boundary can be lowered. Therefore, the magnetization fluctuation of the fixed magnetization layer due to the external magnetic boundary can be suppressed, and the stability of the MTJ element can be improved. Further, since the magnetic flux leaked from the magnetization fixed layer 3a and the magnetization reference layer cancel each other, the magnetic flux leaking from the fixed magnetization layer can be suppressed to the minimum by adjusting the film thicknesses. .3⁄4 layer 5 is composed of a strong magnetic conductor having uniaxial magnetic anisotropy, which can change the magnetization direction by external action, and can maintain the magnetization direction as a beryllium. For example, the magnetization direction is for the magnetization reference layer 3c. The magnetization direction is "parallel" or "anti-parallel", respectively, as "〇" and "1". Between the magnetization reference layer 3c and the memory layer 5, a channel insulating layer 4 of a non-magnetic thin insulating layer is provided, and the magnetic layer is formed by the magnetization reference layer 3c' channel 137416.doc •18·201003651 insulating layer 4 and memory layer 5. Channel engagement (and). On the other hand, in the semiconductor substrate 1 of the lower portion of the cell, the substrate is cut, and the well region is separated from the well region, and is set as the selection transistor 1 for selecting the memory cell. There is a MOS type field effect transistor composed of a closed-electrode insulating film port, a source electrode 13 , a source region 14 , a two-phase source, an electrode 15 , a drain region 16 and a drain electrode 17 . The gate electrode 15 of the above-mentioned 'selective transistor 1G' is formed by connecting cells to each other to form a strip shape, and also serves as a word line of the first column wiring. Further, the electrode electrode 17 is connected to the column wiring 19 as the second column wiring, and the source electrode 13 is connected to the base layer 1 of the spin injection bridging element 2 via the connection plug 7, and The protective layer 6 of the AMTJ element 2 is connected to the bit line 18 which is provided on the upper portion of the memory cell as a row wiring. ° When the information is recorded on the spin-filled component 2Q of the desired memory cell, a selection signal is applied to the word line 15 including the column of the desired memory cell, so that the selection transistor 1 of the column is all Turn on (on) state. : 匕 ; 匕 3 has the desired § 己 体 体 体 体 体 体 体 体 体 与 与 与 与 列 列 列 列 列 列 列 列 列 列 列 列 列As a result, the desired memory cell is selected, and the turbulent flow is applied to the spin-injecting MTj element plus the memory layer 5, and the memory layer 5 is magnetized in a specific magnetization direction for information recording. At this time, first, the magnetization of the MTJ element 2 is spin-injected, and the direction is "anti-parallel" to the magnetization direction of the memory layer 5. The hunting is reversed to the magnetization direction of the memory layer 5 for the magnetization reference layer. When the magnetization direction is "parallel", the write current of the current density above the limit value of the ^ "" limit value flows from the memory layer 5 to the magnetization base 137416.doc 19 201003651 quasi-layer 3c. By this, as a solid, the spin-polar electron flow of the electron density above the threshold will flow from the magnetization reference layer 3c to the memory layer 5, causing magnetization reversal. Conversely, when the magnetization direction of the magnetization reference layer 3c in which the magnetization direction of the memory layer 5 is in the "parallel" state is reversed to the "anti-parallel" state, the current density of the current density above the threshold value is applied to the above. In the opposite direction, that is, from the magnetization reference layer 3c to the memory layer 5, as an entity, an electron flow of electron density above the threshold flows from the memory layer 5 to the magnetization reference layer 3c. The information reading from the 疋/master MTJ element 20 is performed using the tmr effect. Both the writing and the reading of the spin injection MTJ element 2 are used. The interaction between the electrons in the hidden layer 5 and the spin-polar current of the cross-flow is performed in a region where the current density of the spin-bias current is small, and the current is applied to the spin-bias current. It is carried out in areas with high density and exceeding the threshold. Further, in order to prevent the magnetization from being reversed or unstable during the recording operation, the magnetization reference layer 3c is combined with an antiferromagnetic magnet such as PtMn or IrMn to fix the magnetization direction, and a material having a large coercive property such as CoPt is used to be processed to be larger than The area of the memory layer 5 may be used or may be magnetized in a specific direction by an external magnetic field. The magnetization reference layer 3c may be magnetically bonded to the magnetization fixed layer 33 as a separate ferromagnetic layer or as an intermediate layer 3b composed of a non-magnetic metal such as b. The magnetization of the magnetization reference layer 3e is in-plane magnetization and vertical magnetization. Further, B ^ ^, the magnetization reference layer 3c is disposed on the side of the memory layer 5, is disposed on the upper side, or is disposed on the upper side. 137416.doc •20· 201003651 The channel insulating layer 4 should be composed of a ceramic material such as oxide or nitride. In particular, if a magnesium oxide Mg layer is provided as the channel insulating layer 4, and a layer is provided on at least the channel insulating layer 4 side of the magnetization reference layer 3c and the sigma layer 5, a large magnetic resistance change rate can be obtained, which is preferable. Fig. 1 is a graph showing an example of a write pulse train which is not based on the recording method of the magnetic memory element of the embodiment. In the implementation type i, when a single resource is recorded, after the main pulse wave, a secondary pulse wave having the same pulse wave height as the main pulse wave and having a pulse width shorter than the main pulse wave is applied. The main pulse and the secondary pulse wave can be voltage control, current control, or power control. Fig. 1 (1) shows a case where a sub-pulse wave is applied after one main pulse wave. The main pulse wave system is the same as the conventional case of writing with a single pulse wave, and is a pulse wave having a pulse wave height and a pulse width sufficient for recording information. In this case, as described above, the actual write A ' of the spin torque MRAM memory chip having a capacity of several hundred Mbits takes into account the deviation of the inversion threshold of the spin injection MTj component or is caused by the transistor and The deviation of the inversion threshold of the wiring, etc., is applied to the write pulse which is much larger than the average value of the inversion threshold. As a result, the higher the write pulse wave is, the higher the write error rate is, the higher the record house error is. In the conventional writing by a single pulse wave, since the writing error caused by writing by the main pulse is not corrected, the result is as it is, and thus the writing error rate is high. Moreover, the tolerance to external magnetic fields during writing is low. On the other hand, in the present embodiment, after the main pulse wave, the sub-pulse wave having the pulse wave height exceeding the inversion threshold value is applied. Therefore, the main pulse wave can be corrected by the writing of the sub-pulse wave. The write errors generated are highly variable. Moreover, 'because the pulse width of the secondary pulse wave is shorter than the pulse width of the main pulse wave, the 137416.doc -21 - 201003651 is erroneously inserted by the secondary pulse wave, and it is difficult to accumulate the above-mentioned high record. In voltage #, it is not easy to pack & wrong. According to the magnetic memory component of the above-mentioned effect type, according to the method of the present invention, the tolerance of the external magnetic field is improved by writing the wrong 莩, 士, and writing 蚪. , minus ^,, at this time, in the main pulse wave too, ,; u between the end of the end and the side of the auxiliary pulse wave, should be π s i-Λ on the 'more official · stop · ° ° set 3 more And sigh 5 times the above interval. This is used to ensure that the time is corrected by the main pulse. 1 (7) is shown in Fig. 1 (7) after the "main pulse" is applied to the main pulse wave, the secondary pulse wave 1 and the secondary pulse wave 2 as the pulse width; A combination of three consecutive pulse waves (four) a legend. In this case, the writing by the field j pulse 1 and the writing by the sub-pulse 2 are repeated twice, and the pulse wave width is shorter as the pulse wave is applied later. A high recording voltage error due to accumulation of excess energy occurs, so the possibility of improvement in the write error rate becomes higher. Fig. 2 is a graph showing an example of a write pulse train according to an embodiment of the present invention, which shows an example of various main pulse waves. Fig. 2(a) and Fig. 2(b) show an example in which a short stop period in which the write pulse power injection is stopped is set in the main pulse wave. As shown in Fig. 2(a), even if a stop period is provided in the middle portion of the main pulse wave, there is no effect. However, if a stop period is provided near the end of the main pulse wave as shown in Fig. 2(b), there is a certain period of time. The write power to be injected is effectively reduced gradually, so that the above-mentioned high recording voltage error is unlikely to occur (refer to Japanese Patent Application No. 2008-107768). 2(c) and 2(d) show an example in which two main pulse waves are applied. Fig. 2(c) shows the application of 137416.doc -22· 201003651 with the main pulse wave and the main pulse wave 2 with equal pulse wave height and pulse width, and Fig. 2(d) shows the application of pulse wave height and pulse width mutual The case of the main pulse wave and the main pulse wave 2 are different. In any case, the writing by the preceding main pulse 无效 is invalidated by the writing by the subsequent main pulse 2, so that the effect of applying the complex main pulse is not particularly effective. Fig. 3 is a graph showing an example of a write pulse train according to an embodiment of the present invention, which shows an example of various auxiliary pulse waves. Fig. 3(a) is an example of applying 2 悃 after the main pulse wave, generally applying a plurality of sub-pulse waves, and repeating the correction twice by using the writing by the sub-pulse wave, generally repeating the plurality of times, so the writing error The possibility of rate improvement becomes higher. In this case, as described above with reference to Fig. i(b), it is expected that the pulse width of the sub-pulse wave is gradually decreased. On the other hand, Fig. 3(b) and Fig. 3(c) are provided with an example of a sub-pulse wave preceding the last main pulse wave, and the effect of applying such a sub-pulse wave is not particularly applied. The implementation type 2 is in the implementation type 2, and is mainly directed to other examples of the recording method for the requesting item 丨 and the spin-in which the MTJ tl piece is generated. Fig. 4 and Fig. 5 are graphs showing an example of a magnetic memory cell according to the second embodiment of the present invention. In the implementation mode 2, when one piece of information is recorded, after the main pulse wave, the auxiliary pulse wave having the same pulse width as the main pulse wave and the skin π being lower than the main pulse wave is applied. The main pulse and the secondary pulse are voltage protection. The flow control or power control can be used. Fig. 4 shows a case where a sub-pulse wave is applied after one main pulse wave, and Fig. 4 (hook (8) is an explanatory diagram for explaining the effect. The main pulse wave system is compared with the case of writing with a single 'wave. $, which is a pulse wave having a pulse wave width of 137416.doc -23- 201003651 sufficient to record information. In this case, as described above, for a spin torque MRAM memory chip having a capacity of several hundred Mbits The actual writing takes into account the deviation of the inversion threshold of the spin injection MTJ component or the deviation of the inversion product limit of the transistor and the cloth, and the application is much larger than the inversion threshold. The average value is written into the pulse wave. As a result, as shown in Fig. 4 (4), the magnetic memory element having the average inversion threshold value is expressed by V _ + + β i and is found by the main pulse wave. Write, the write power injection is excessive than the reverse margin, and the write error rate is increased by the write error rate. On the other hand, the pulse wave height of the pulse wave is higher than the average reverse S product. Limit, but not {; Bu JJL b 4 -- the degree of reversal of the non-existing reversal threshold is significantly higher Therefore, if the secondary pulse wave is applied, the cold y_ "into" is corrected by the write generated by the main pulse. Moreover, since the energy is left by the auxiliary pulse wave, it is difficult to write the upper and lower lamps. Less injected into the 旲 旲 阿 阿 阿 阿 阿 阿 阿 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 The tolerance of the magnetic field of the mouth p is improved. On the other hand, as shown in Fig. 4(b), in the component element of the inversion threshold, there will be a pulse wave of the auxiliary pulse wave, and ^^ is less than Inversion of the threshold value. §Hei magnetic S recall element, even if the auxiliary pulse wave is applied, it will not be invalid in the auxiliary pulse wave 2, and it will remain written by the main pulse wave as it is. The magnetic record is higher than the limit value. However, the "higher than the reversal threshold", the pulse wave height of the main pulse wave is not the value of the private degree, due to the presence of excess energy due to the main pulse wave There is very little error in the write error rate due to the injection, that is, the write error rate by the main pulse wave is small = the record is not corrected. Good write, 137416.doc •24· 201003651 From the above results, if the write pulse train shown in Fig. 4 is used, whether for a magnetic memory element having an average reversal threshold, or Reversing the magnetic memory element # with a high threshold, and writing a good write error rate is small. • Figure 5 shows that two sub-pulses are applied after one main pulse, corresponding to the request • Item 2 The main pulse wave, the auxiliary pulse wave 丨 and the auxiliary pulse wave 2 are combined as a combination of three consecutive pulse waves whose pulse wave height is gradually reduced. In this case, as shown in FIG. 5, for the average inverse The correction of the magnetic memory element having the transition value by the writing by the sub-pulse 1 has the same effect as the case shown in FIG. 4(a). The secondary pulse wave 2 is invalid. As shown in Fig. 5(b), the magnetic memory element having a high inversion threshold is the same as the case shown in Fig. 4(b), and the pulse wave height of the main pulse wave is not significantly higher than the inversion threshold. With good writing by the main pulse, there is no need for correction. Further, as shown in FIG. 5(c), the magnetic memory element having a low inversion threshold is repeated by writing by the sub-pulse 1 and writing by the sub-pulse 2 Two times Cj and the later she added the pulse wave, the more difficult the recording voltage error caused by the excess energy injection, the higher the possibility of improving the write error rate. From the above results, "If the write pulse train shown in Fig. 5 is used, it is possible to perform a better write with a smaller write error rate than in the case of the write pulse train shown in Fig. 4. . [Embodiment] In the embodiment, the spin torque MRAM including the spin injection MTJ element is applied to the recording method according to the embodiment of the present invention, and the method of the present invention is verified by 137416.doc -25-201003651 effect. Examples 1 and 2 are tests for requesting the roots of ws 3 to 5, and Example 4 is an experiment for the basis of claim 2. Each male system applies a magnetic field to the long axis of the component, repeats the erasing & recording, and records the write error rate. The direction in which the magnetic field is applied is in the opposite direction to the direction of magnetization to be recorded. Human Embodiment 1 In Embodiment 1 'corresponding to the recording method of the magnetic memory element according to the implementation type, the write pulse train shown in Fig. 1 (" is applied. The spin torque MRAM used includes the lower The spin formed by the layer is implanted into the mtj element 20. The base layer 1 : a Ta film having a film thickness of 5 nm; the antiferromagnetic layer 2: a PtMn film having a film thickness of 30 nm; the magnetization fixed layer 3 a : a CoFe film having a film thickness of 2 nm Intermediate layer 3b: Ru film with film thickness of 0·7 nm; magnetization reference layer 3 c: CoFeB film with film thickness of 2 nm; channel insulating layer 4: magnesium oxide MgO film with film thickness of 0.8 nm; memory layer 5: film thickness 3 The CoFeB film of nm; the protective layer 6: the film of the Ta film with a thickness of 5 nm is injected into the MTJ element 20 in a plan view having a long axis length of 150 to 250 and a short axis length of 70 to 85 nm, and the magnetization of the memory layer 5 is 140 Oe—the external magnetic field applied to the spin injection MTJ element 20 by 50 〇e is connected to a main pulse wave with a pulse voltage of 0.8 V and a pulse width of 30 ns, and a pulse wave voltage of 88 V is applied. The pulse wave of the width W. At this time, see the pulse wave of the secondary pulse wave and the pulse wave between the end of the main pulse wave and the front end of the auxiliary pulse wave 137416.doc -26- 2 01003651 Various changes are made to the interval D, and the relationship between the write error rate and the write error rate is investigated. Fig. 6 shows the write error rate and the pulse wave interval in the case where the pulse wave width % is applied to the side pulse wave of 丨~3〇ns. A graph of the relationship of 0. Two different trends can be seen from Fig. 6. That is, in the case where the pulse wave is torn as a pulse wave of 1 μ as the auxiliary pulse wave, the pulse wave interval D is 1 ns. In the case of the case, the error rate is changed. The good effect is remarkable. If the pulse interval D exceeds 5 ns, there is almost no improvement. On the other hand, when the pulse wave width is 2 or 3 μ as the secondary pulse wave, the pulse wave is used as the secondary pulse wave. When the pulse wave interval D is 3 ns or more and more desirably 5 ns or more, the error rate improvement effect of the present invention is remarkable. When the pulse wave width W is 5 ns or more as the auxiliary pulse wave, the pulse wave width W is 5 ns or more. In the case where the improvement effect is small, the pulse width w of the right side pulse wave is the same as the pulse width of the main pulse wave and is 30 ns, and there is no improvement at all. Example 2 Example 2 corresponds to the embodiment according to the embodiment. The recording method of the magnetic memory element of 2 applies the write pulse train shown in Fig. 4. The spin force used. The matrix MRAM has the same layer constitution as the spin injection MTJ element 20 used in the first embodiment, and includes a spin-in/simplification element 2G in which the memory layer 5 has a retentive magnetization of 125 〇e. - The surface is from the mMTJ. The element 20 is applied with 5G 〇e. The magnetic field is connected to a pulse wave voltage of 0.9 V and a pulse wave width of 30 ns. The pulse wave voltage V is applied and the pulse wave width is 30 ns. At this time, the pulse wave a £ V of the sub-pulse wave and the pulse wave interval D between the end of the main pulse wave and the front end of the sub-pulse wave are variously changed, and the relationship with the write error rate is investigated. Fig. 7 shows that the pulse interval D is changed within a range of 1 ns, and a 137416.doc -27-201003651 surface = the error rate of the human error and the pulse wave voltage ratio of the main pulse and the secondary pulse wave. The graph. Although not as clear as shown in Fig. 6, in Fig. 7, it is considered that there are two different trends. $ is the pulse interval D is set to 3 ns in the case of 卜 ̄ g -w τ M a- r j, in the case of the gate ns M, only the main pulse and the vice

When the pulse wave voltage ratio of the pulse wave is D 7 L or more and 0.7 or less, the improvement effect is observed. Especially when the pulse wave power ratio is 0 8 or more and 〇·95 or less, the temple improvement effect is remarkable. The pulse wave voltage of the effective side pulse wave has a lower limit, indicating that the writing by the side pulse wave is being performed. In the case where the pulse wave interval D is set to 1 ns or 2 ns, the improvement effect of the pulse wave voltage ratio of the main pulse wave and the auxiliary pulse wave is G_8 or more and 〇.95 or less, and it is judged that In the same way, according to the invention, the pulse wave voltage ratio is 〇·3 or more, Q 95 or less, and the pulse wave electric delay J of the auxiliary pulse wave is in the case of the inverted L limit. The effect of his invention. [Embodiment 3] In Example 3, the tolerance to external magnetic fields of the recording method of the magnetic memory element according to the embodiment was investigated. The spin torque mram used has the same layer structure as that of the spin injection MTJ element 2A used in the first embodiment, and includes the spin injection MTJ element 20 in which the repulsive layer 5 has a coercivity of 2 丨 2 〇e. Write error rate in the case where the spin pulse is injected to apply the external magnetic field of 〇~2〇〇〇e to the element 2〇 while the write pulse voltage is changed within the range of 〇5 to 〇7 V. . The polarity of the voltage is positive. 8 is a diagram showing the external magnetic field connected by a contour line, and the writing error rate is expressed as a position of 〇丨, 〇〇1, and 〇〇〇1 with respect to the write pulse voltage, respectively. 137416.doc •28· 201003651 埘If the stone becomes larger in the field, then it must fight against the powerful external magnetic field to write information more. Therefore, in order to maintain the same write error rate, the pulse voltage must be formed. Therefore, it is predicted that the above contour line will have a greater life pressure in Fig. 8 = line. Further, if the external magnetic field is constant, the larger the predicted pulse wave energy i, the smaller the intrusion error rate becomes. == Comparative example of single-pulse recording with a pulse width of 100 ns y, , - mouth fruit. In this case, the external magnetic field: as predicted, becomes the curve of the right rise, but after the external magnetic second, the second cover is empty, and even if the pulse voltage is increased, the write error rate is still not the same. In this area, when the external magnetic field is fixed, the pulse wave is large, and the error rate of the writing person becomes larger, that is, the high recording power is caused to be the other side. FIG. 8(4) is after the main pulse wave having a pulse width of 100 ns. After setting the pulse interval of 3 ns, the pulse wave width of 3 ns is applied to the area where the π magnetic field is large, and the contour line becomes right up. If the external magnetic field is constant, the pulse voltage is larger, and the two are written. :small. Thus, according to the recording method of the first embodiment, the error rate of the writer is improved, and the range of the error can be increased, and the tolerance to the magnetic % is enhanced by the action of a strong external magnetic field. = If this is based on the recording operation of the magnetic memory component according to the implementation type, the magnetic recording of the external magnetic field can be shielded in the environment;!, can be reduced - miniaturized, lightweight Spin torque J37416.doc -29- 201003651 Embodiment 4 In the fourth embodiment, the investigation is used as the write pulse train, and the main pulse wave combined with the pulse width 丨〇 ns has various pulse widths and pulse waves. The write error rate in the case of the pulse wave of the secondary pulse wave. At this time, corresponding to the implementation type 丨, the pulse waves of the main bifurcation wave and the auxiliary pulse wave are still set to be the same, and the pulse width of the secondary pulse wave applied later is the same as the pulse width of the previously applied auxiliary pulse wave, or Shorter than it. The spin torque MRAM used has the same layer composition as the spin injection MTJ element 20 used in the embodiment ,, and the white magnetic ^ 曰 取 包 3 有 有 有 有 有 己 己 己 己 己 己 己 己 己

The spin of Oe is injected into the MTJ element 20. On one side of the 兮 妒, + ® 4 spin > the main MT J element 2 0 to make the external magnetic field of 50 〇e, the surface application, the main pulse wave and the auxiliary pulse wave with a pulse voltage of 1.1 V. 1 is called λγ Μ sequence, indicating the pulse width and pulse interval of the main pulse and the secondary pulse, and finally indicates the write error rate in the case of using the write pulse train. [Table 1 ]

137416.doc Secondary pulse 2 (ns) Pulse interval 3 (ns) Secondary pulse 3 (ns) Write error rate (%) 4.2xl〇'3 1.3xl0'3 2 2.8xl0'5 1 1 1 7.6xl 〇·4 2 1 1 1.8xl〇·5 8.0χ1〇·2 8.1χ1〇-2 Ι.ΙχΙΟ-3 2.2χ1〇'3 1 1·7χ1〇-4 •30. 201003651 Comparative Example 1 is applied with a single pulse In the case of waves, the write error rate in this case is 8.0 χ 10·2. Comparative Example 2 is a case where a secondary pulse wave is applied first to the main pulse wave. In this case, the write error rate is 81 χ 1 〇 -2, which is constant within the error range from Comparative Example 1, indicating that the secondary pulse wave preceding the main pulse wave is invalid. The pulse train 1 and the pulse train 2 are applied to the main pulse wave 10 (10), and the sub-pulse is applied. In this case, the pulse width of the sub-pulse is 3 ns better than 2 ns. This is because the pulse width may be slightly too short if 2 ns is used for sufficient writing by the secondary pulse. When the pulse trains 3 to 5 are connected to the main pulse wave and two or three sub-pulse waves are applied, the write error rate is improved as compared with the pulse wave train 2 in which the sub-pulse wave is fixed. Except for the secondary pulse wave which is invalid before the pulse wave train 6, the pulse wave after the main pulse wave is the same as the pulse wave train 2, and the write error rate is also substantially the same. The pulse train 7 and the pulse train 8 are applied to the short pulse wave having a pulse width of 1 ns after 1 ns of the main pulse wave. The write error rate is improved as compared with Comparative Example 1. In this case, the pulse wave train 8 of the short sub-pulse wave having two pulse widths of 1 ns is one pulse wave train 7 of the sub-pulse, and the write error rate is further improved. Here, as described above, the results of the pulse train 7 and the pulse train 8 may overlap the effect of the bite of the present invention and other inventions. Next, a circuit for generating a write pulse wave of the above embodiment will be described. Fig. 13 is a view showing a configuration of a write pulse wave generating circuit for generating an address pulse of a main pulse and a sub-pulse in the above-described embodiment from one rectangular pulse wave. The write pulse wave generating circuit 30 is composed of a plurality of buffers 3 2, 3 3, "4" and a plurality of logic circuits 35 and 36. The wheel pulse wave generating circuit 3 is turned on. 137416.doc -31- 201003651 The end 3 1 input has a rectangular pulse wave signal. The rectangular pulse signal input to the input terminal 31 is input to one of the input terminals of the logic circuit 36 of the OR gate logic, the non-inverting input terminal of the logic circuit 35 of the gate logic, and the buffer 32 connected in series. 33. Here, the buffers 32, 33 connected in series are used to generate the width of the sub-pulse, and the time tdl of the arbitrary sub-pulse can be selected by the selection of the delay times of the buffers 32, 33. The outputs of the buffers 32, 33 are input to the non-inverting input of the logic circuit 35 of the AND gate logic. The output of the logic circuit 35 of the AND gate logic is input through buffer 34 to the other input of the logic circuit 36 of the OR gate logic. Here, the buffer 34 generates the time td2' between the main pulse wave and the sub-pulse wave by the delay time of the buffer 34, and can set an arbitrary time td2. Then, the write pulse wave composed of the main pulse wave and the sub pulse wave is obtained by the logic circuit 36 of the OR gate logic, and is output from the output terminal 37 of the write pulse wave generating circuit 30. Fig. 14 is a view showing the configuration of the write pulse wave generating circuit 4A of the pulse wave generated by the waveform memory and the D/A conversion circuit. The waveform is stored in a waveform of a write pulse composed of a main pulse and a sub-pulse. The waveform data of the wide pulse wave can be selected from the 2n stage to select the N bit of the output bit as the old %' to form the time series data of the complex digital element. The waveform memory 4 is provided with u 珲 for reading, and the input terminals of the (10) ( (4) conversion electric power (4) are respectively connected. d/a conversion circuit (4) From the waveform § Remembrance 41, the waveform data that is driven into the clear, and + into the pulse wave is input into the material of each N-bit (1 character) and converted into analog-to-digital conversion (4) For example, a ladder is used as a pulse wave output such as a ladder resistor circuit. By writing such a pulse wave generating circuit, the gate pulse is obtained by the write pulse wave 137416.doc -32· 201003651, and the write pulse of each of the above embodiments can be obtained easily and with a high degree of freedom. Further, in the example of Fig. 8, in order to determine the output level from the 23rd stage, the number N of the 1-character bits is set to "3", but the present invention is not limited thereto. The present invention has been described above on the basis of the embodiments, but the present invention is not limited thereto, and may of course be appropriately changed without departing from the spirit of the invention. [Industrial Applicability] According to the present invention, it is possible to improve the transient characteristics at the time of writing, and the writing failure is small, the threshold of the writing current density is small, and the integration, the speed, and the low consumption can be achieved. The spin-injected magnetization inversion type mtj device for electric power can contribute to the practical use of small, light, and low-priced non-volatile memory. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a graph showing an example of a write pulse train of a magnetic memory element according to an embodiment of the present invention; FIG. 2(a) to (d) A graph showing an example of a write pulse train of a recording method of the same magnetic memory element; and FIGS. 3(a) to 3(c) are diagrams showing an example of a write pulse train of the recording method of the same magnetic memory element. 4(a) and 4(b) are graphs showing an example of a write pulse train of a recording method of several magnetic memory elements according to an embodiment 2 of the present invention; Fig. 5(a) ~(c) is a graph showing an example of a write pulse train of the recording method of the same magnetic memory element; 137416.doc -33-201003651, and Fig. 6 shows a magnetic memory element of the embodiment 1 of the present invention. A graph of the relationship between the error rate and the pulse interval of the recording side; and FIG. 7 shows the writing error rate and the height of the auxiliary pulse wave of the recording method of the magnetic memory element of the second embodiment of the present invention. A graph of the relationship; - Fig. 8(a) 8(b) shows a magnetic memory according to the third embodiment of the present invention. The graph of the tolerance of the recording to the external magnetic field; Figure ^ is the table *MT; the basic structure of the component and the reading of the memory information. (a) 'and the mram composed of MTJ components A partial perspective view of a structure of a memory ugly cell (b); a partial perspective view showing a structure of a spin torque MRAM shown in Patent Document 2; and FIG. 11 is a view showing a same composition of a spin note. Spin torque MRAM^ ^ ^ ^ ^ ^ ^ ^ ^ ^ . Figure 12 is a graph showing the relationship between the write pulse voltage and the write error rate; Figure 13 shows the slave! The rectangular pulse wave generates a configuration of the write pulse wave generating circuit including the write pulse wave of the skin and the d pulse wave in the above embodiment; 'Fig. 14 shows that the waveform memory and the d/a conversion circuit are not used. FIG. 15 is a partial perspective view showing a structure of a memory cell of a spin torque mram according to an embodiment of the present invention; and FIG. 15 is a partial perspective view showing a structure of a memory cell of a spin torque mram according to an embodiment of the present invention; A cross-sectional view showing the configuration of a spin injection element of an embodiment of the invention. 137416.doc -34- 201003651 [Description of main component symbols] 1 Base layer 2 3 a 3b 3c 4 5 6 7 10 11 11a 12 13 14 15 16 17 18 19 20 21 30, 40

Antiferromagnetic layer magnetization fixed layer intermediate layer magnetization reference layer channel insulation layer memory layer protection layer connection plug selection transistor semiconductor substrate well region gate insulating film source electrode source region gate electrode > and polar region bungee Electrode bit line column wiring spin injection magnetization inversion MTJ element element separation structure write pulse wave generation circuit 137416.doc -35-

Claims (1)

201003651 VII. Patent application scope: 1. A magnetic memory component recording method, the magnetic memory component at least comprising: a memory layer comprising a strong magnetic conductor, the magnetization direction can be changed, and the information is used as a magnetic circuit And a reference magnetization layer, wherein the reference magnetization layer is provided for the memory layer isolation layer, comprising a strong magnetic conductor, the magnetization direction being fixed and being the reference of the magnetization direction; and flowing through the insulating layer Recording the current between the memory layer and the reference magnetization layer; when recording one piece of information, applying more than m of the main pulse wave of more than m to the same direction; and one or more of the main pulse waves Thereafter, one or more of the aforementioned pulse waves are applied; 7 the auxiliary pulse wave applied to the main pulse wave is a pulse wave having a pulse width wider than the main pulse wave, or a pulse wave having a lower pulse wave than the main pulse wave The pulse of at least one condition of the wave. 2. The method of recording a magnetic memory device according to claim 1, wherein at least one of the pulse waveforms including the one or more main pulse waves of the month J and the one or more of the auxiliary pulse waves applied thereafter is set. A combination of three consecutive pulse waves and a combination of at least one of a pulse width and a pulse height is gradually reduced. 3. The method of recording a magnetic memory device according to claim 1, wherein a time of 3 ns or more is set between an end of the main pulse wave of the above i or more and a front end of the one or more of the sub-pulse waves applied later. Interval; (wherein, the end of the pulse wave and the front end are respectively the drop of the pulse wave and the position where the Ada is one of the maximum values of the pulse wave height; the same applies hereinafter). The method for recording a magnetic memory device according to claim 1 or 2, wherein the one or more main pulse waves and the pulse wave train of the one or more of the sub-pulse waves applied thereafter are included in In a combination of two consecutively selected pulse waves, the posterior pulse wave is at least one condition that the pulse width is 2 ns or more and 1 〇 (10) or less, or the pulse wave height is 0.7 times or more and 95 times or less of the front pulse wave. The pulse wave is set at a time interval of 5 w or more between the end of the pulse wave and the front end of the back pulse wave. 5. 6. The recording method of the magnetic memory element according to the π item 1 or 2, wherein the one or more main pulse waves and the pulse wave train of the one or more of the pulsation waves applied later are included 5# The selected posterior pulse wave of the continuous 2 pulse waves corresponds to a pulse width of 3 ns or less, or the pulse wave height is 0.8 times or less of the pre-pulse wave s + I main v - condition 'and the precursor The time interval between the end of the wave and the front end of the back pulse is less than 5 ns. The method for recording a magnetic memory element of the 1st or 2nd member, wherein the one or more main day ρΏ u _ pulse waves and the one or more of the previous ones before the pulse wave are applied In the group of two consecutive continuous pulse waves, the H-shaped pulse width is 3 ns or less, or the pulse wave height is 0.95 times of the front pulse wave to the v-condition, and the pre-pulse wave is obtained. The time interval between the end of the end and the front end of the back pulse is less than 5 ns. 137416.doc